Regulatory Landscape Briefing Bot

A Slack or web bot that answers ‘what does current AI regulation say about X?’ by grounding answers in tracked legislative texts across jurisdictions.

Difficulty: 1-week | Stack: Python, LangChain or LlamaIndex, pgvector + PostgreSQL, FastAPI, Slack Bolt SDK, Next.js (optional web UI)

Who this is for

Developers and product managers at AI startups who need quick, cited answers about compliance obligations (EU AI Act, US executive orders, UK AI framework) without hiring a policy team.

Build steps

1. Collect and chunk key regulatory documents: EU AI Act (full text), US EO 14110 and NIST AI RMF, UK AI Safety Institute guidance, and any state-level bills (e.g. California SB 1047 successor drafts)
2. Embed all chunks into pgvector with metadata: jurisdiction, document name, section number, publish date
3. Build a RAG pipeline that retrieves the top-k relevant chunks for a user query and prompts an LLM to synthesize an answer with inline citations back to specific articles/sections
4. Wire the pipeline to a Slack Bolt app so teams can query it inline (e.g. ‘/aipolicy Does the EU AI Act require incident reporting for general-purpose models?’)
5. Add a freshness layer: track document versions and alert the Slack channel when a source document is updated so users know their answers may be stale

Risks

• Regulatory texts are dense and cross-referential — naive chunking loses context between articles, producing confident but incomplete answers; use hierarchical or parent-document chunking strategies
• Legal language is jurisdiction-specific and the bot may conflate requirements across regions if the metadata filtering isn’t enforced in retrieval — always surface the jurisdiction in the cited answer
• Keeping the document corpus current is an ongoing manual burden; without a clear update workflow the bot quietly becomes outdated, which is worse than no bot at all